Johan G.L. GL Verheyden

Johan G.L. GL Verheyden

orcid

Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 July 2026 in medRxiv
Verified
Preprint
Open access
PDF file
JVJohan G.L. GL Verheyden
Johan G.L. GL Verheyden
CMCelestin Nzanzu Mudogo
Celestin Nzanzu Mudogo
WJWolfgang Jacquet
Wolfgang Jacquet
Background: The 2026 outbreak of Bundibugyo virus disease in the Democratic Republic of the Congo generated the largest recorded epidemic caused by Bundibugyo virus and displayed unusually rapid early growth. Early scenario projections warned that the outbreak could become one of the largest Ebola-family epidemics on record, but were calibrated to assumed mortality totals and intervention scenarios rather than evaluated repeatedly against subsequently observed daily surveillance data. We assessed how accurately the outbreak could have been forecast in real time using routinely published national situation reports, whether a simple ensemble of two complementary models improved on either alone, and how the resulting forecasts and their uncertainty should be interpreted for operational capacity planning. Methods: We conducted a retrospective, pseudo-prospective rolling-origin forecasting study using daily cumulative confirmed cases and deaths reported in national situation reports from 14 May to 20 July 2026 (68 calendar days; 61 numeric national reports). Every date with a newly reported national total became an eligible forecast origin once at least seven numeric observations were available (55 origins). At each origin, all later observations were withheld and two prespecified operational models were refitted using only data available at that time: a recent seven-increment baseline and a Bayesian negative-binomial surveillance-maturity model fitted by full Markov Chain Monte Carlo on complete available history. A Gompertz growth model was also fitted identically at every origin as a phenomenological benchmark. Forecasts were generated for 7-, 14-, and 21-day horizons and evaluated only when an observation existed on the exact target date. The Bayesian model’s forecasts were additionally corrected by a prospective, quality-gated empirical recalibration using only errors already realised at earlier, forecast-maturity-qualified origins. We further evaluated a data-driven ensemble that pooled the baseline’s bootstrap trajectories with the recalibrated Bayesian model’s posterior predictive trajectories, with horizon- and outcome-specific mixing weights selected by a strictly prospective, expanding-origin procedure using an explicit asymmetric operational loss function. Findings: A statistical and externally corroborated surveillance-maturity discontinuity was identified on 28 May 2026 (robust z-score 13·0 against the trailing week). The Gompertz model had the poorest 80% predictive-interval coverage of all candidates at every horizon and outcome (25·9–34·1% for cases, 0·0–23·5% for deaths) and is reported as tested and excluded. Recalibration corrected a directionally consistent Bayesian underprediction tendency and is the primary reported Bayesian result; the correction was applied at 24 of 55 origins for the 21-day horizon (never at 28 days in earlier work), because 21-day forecasts reach the forecast-maturity threshold (mature history ≥ forecast horizon) from the 18 June origin onward, a genuinely more useful operational horizon than 28 days. Recalibration improved case coverage substantially (7-day 80% coverage 60·0%→80·0%; 14-day 41·0%→76·9%) and improved death forecasts on both accuracy and calibration (7-day median absolute percentage error 10·3%→7·7%; 80% coverage 33·3%→95·6%). The baseline–Bayesian ensemble’s evidence-selected weighting differed materially by outcome: for cases, the operationally and distributionally optimal blend was baseline-heavy (w≈0·8–0·9 toward the baseline) at every horizon; for deaths, the optimal blend was close to parity (w≈0·5–0·6), and the ensemble improved mean weighted interval score over both individual component models at every horizon (e.g., 7-day: baseline 24·8, Bayesian 20·5, ensemble 17·9). Interpretation: Routine SitRep data supported useful short-term forecasting, but no single model was best on every criterion, and a data-driven combination of two complementary, individually weaker-in-some-respect models measurably outperformed either alone for one of two outcomes evaluated. Three distinct kinds of maturity determine whether a forecast at a given horizon can be trusted: epidemic maturity, surveillance maturity, and forecast maturity; the last of these is the strongest and most model-agnostic finding in this paper, and motivates reporting 21- rather than 28-day forecasts as the longest routine operational horizon. We report an explicit, reproducible, and adjustable framework — not a fixed recommendation — for combining simple and complex models and for communicating the resulting uncertainty to non-specialist operational decision-makers. Funding: none received specifically for this analysis. Author Summary: When an outbreak is growing, planners need to know not just how many cases to expect, but how much to trust that number. We tested, in a fully retrospective and fair way, how well three forecasting methods — a simple recent-trend method, a saturating growth curve, and a more sophisticated probabilistic model — could have predicted the 2026 Bundibugyo virus disease outbreak in the Democratic Republic of the Congo one, two, and three weeks ahead, using only the information that would genuinely have been available at each point in time. The saturating growth-curve model performed worst on every measure and is not recommended. The simple trend method gave the most accurate single-number forecasts for confirmed cases, but its stated uncertainty ranges were frequently wrong. The probabilistic model’s ranges were much better calibrated once we corrected a systematic bias using only past, already-known forecast errors — a technique we describe in enough detail that others can reuse it on their own outbreaks. Combining...

30 July 2026 in medRxiv
Verified
Preprint
Open access
PDF file
JVJohan G.L. GL Verheyden
Johan G.L. GL Verheyden
CMCelestin Nzanzu Mudogo
Celestin Nzanzu Mudogo
WJWolfgang Jacquet
Wolfgang Jacquet
Dashboards and situation reports publish deaths per confirmed case by health zone, and decisions on investigation and resources lean on these numbers. We ask when such a reported ratio is ready to be read as fatality, and propose a readiness test that combines Bayesian partial pooling, an ascertainment tipping-point analysis and repeated refitting at successive data cut-offs, framed by a surveillance-maturity logic. We applied it to public cumulative confirmed cases and deaths from the 2026 Bundibugyo virus disease epidemic in the Democratic Republic of the Congo (8,728 cases and 4,205 deaths on 6 October). Among 41 health zones with at least 10 cases, North Kivu zones had higher adjusted odds of a reported death than Ituri zones (odds ratio 1.46 (0.99-2.14), 95% credible interval), a point estimate that was insensitive to the prior, the case threshold and omission of single zones. The estimate was nevertheless unstable across cut-offs: the odds ratio was 1.78, 2.54, 1.67, 1.51 and 1.46 on 20 July, 30 August, 30 September, 5 October and 6 October (the last three from identical zones), and its lower credible limit, 1.01 on 5 October, was 0.99 on 6 October. A 1.25-fold difference in the ratio of death to case ascertainment between the provinces would suffice to produce the contrast. Cases and deaths were often reported on the same date, and mapped facility indicators added nothing. By this test, zone-level ratios are not yet interpretable as fatality, although they correctly show where to investigate. In simulated stable epidemics the cumulative decline since 25 August was never matched, whereas the drift ratio over three cut-offs was a weak signal. Surveillance-design actions follow: date and label ratios, show uncertainty and drift, record event dates, audit ascertainment, and link patient records.

3 August 2026 in medRxiv
Verified
Preprint
Open access
PDF file
JVJohan G.L. GL Verheyden
Johan G.L. GL Verheyden
CMCelestin Nzanzu Mudogo
Celestin Nzanzu Mudogo
WJWolfgang Jacquet
Wolfgang Jacquet
Background: A retrospective field investigation described substantial illness and mortality in and around Mongbwalu from January 2026, several months before Bundibugyo virus disease (BVD) was formally recognised in May. Historical BDBV evidence includes a genomic reconstruction compatible with multiple spillovers, while other orthoebolavirus outbreaks have involved survivor-associated resurgence, concurrent lineages and substantial syndromic misclassification. We evaluated whether the early Mongbwalu signal was most compatible with continuous acute transmission, a later introduction, survivor-mediated reseeding, or mixed/non-ancestral mortality. Methods: We combined a structured rapid review of 35 sources with official confirmed surveillance, a published report of an unpublished retrospective investigation, and a soft genomic timing constraint. A resample-move approximate Bayesian computation sequential Monte Carlo model simulated 20 weekly periods across Mongbwalu, Bunia, Rwampara and Nizi. Four hypotheses were compared: continuous acute transmission (H1), repeated zoonotic introduction or temporarily elevated shared exposure (H2), survivor-mediated sexual transmission from persistent virus in semen (H3), and mixed or substantially non-ancestral retrospective mortality (H4). Three independent chains of 450 particles were run for each hypothesis over seven decreasing tolerances. Model recovery, prior sensitivity, posterior-predictive checks and leave-one-component-out analyses assessed identifiability and robustness. Results: Under equal priors, H4 received 36.2%, H2 32.6%, H3 21.6% and H1 9.6%. Under the literature-neutral prior, H2 received 50.0%, H4 34.7%, H1 11.1% and H3 4.2%. H2 and H4 jointly accounted for 84.8% under the main prior and 68.7%-89.3% across all prior families, demonstrating a robust preference for a discontinuous origin history. The approximate evidence ratio was only 1.11:1 for H4 over H2, so H2's lead under the main analysis reflected both close empirical fit and direct BDBV precedent incorporated into the prior. H1 was correctly recovered in 92% of equal-prior synthetic datasets, compared with 52% for H2, 42% for H3 and 24% for H4. Removing confirmed May counts eliminated H2's clear advantage, identifying the rapid May expansion as the principal evidence for a later successful seed. Interpretation: The findings favour a discontinuous origin: the January/February Mongbwalu signal was epidemiologically meaningful, but the lineage that expanded in May most probably arose from a later successful introduction or renewed primary exposure rather than from one uninterrupted acute chain. H2 is the preferred literature-informed explanation because it best reconciles the structured early signal, March-centred sampled ancestry and rapid May growth, while direct BDBV precedent makes the mechanism credible. H4 remains a substantial alternative because the early events were not laboratory confirmed and H2 and H4 are only partly identifiable. H3 is temporally and biologically possible but lacks survivor-specific evidence, and H1 is not the leading explanation under any tested prior family.

12 August 2026 in medRxiv
Verified
Preprint
Open access
PDF file
JVJohan G.L. GL Verheyden
Johan G.L. GL Verheyden
CMCelestin Nzanzu Mudogo
Celestin Nzanzu Mudogo
Anticipating which health zone will report the next confirmed case is operationally distinct from forecasting national case counts and matters for prepositioning response capacity; most spatial spread models rely on mobile-phone mobility data unavailable in the Democratic Republic of the Congo (DRC). We modelled the discrete-time hazard of a first reported confirmed case across 106 health zones in four provinces affected by the 2026 Bundibugyo virus disease outbreak (47 affected, 59 at risk, 26 July 2026), comparing four connectivity specifications — none, road-distance, a gravity score, and an incidence-weighted force-of-infection (FOI) term — fitted within an identical Bayesian hierarchical hazard architecture. Evaluation used a rolling-origin design, cluster bootstrap resampling, leave-one-origin-out and non-overlapping-origin checks, and a kernel-parameter sensitivity grid, with top-10 hit rate the pre-specified primary metric, matched to the operational question of which few zones warrant attention; AUC-PR, top-5 hit rate, and median rank percentile were secondary. FOI had the highest top-10 hit rate (42.6%), approaching conventional significance against road-distance and no-connectivity comparators. On AUC-PR, a model with no connectivity term performed as well as or better than any connectivity specification (0.437 vs. 0.409 for FOI), a discrepancy we report rather than omit. Rankings were stable across the sensitivity grid (Spearman ρ 0.90–0.99) and across robustness checks. An incidence-weighted connectivity term modestly and specifically improves identification of the highest-risk zones, concentrated in top-k ranking rather than uniform across metrics. The evaluation is pseudo-prospective, since historical data-vintage snapshots could not rule out retrospective revision, pending verification via a pre-registered top-20 ranking.

14 August 2026 in medRxiv
Verified
Preprint
Open access
PDF file
JVJohan G.L. GL Verheyden
Johan G.L. GL Verheyden
CMCelestin Nzanzu Mudogo
Celestin Nzanzu Mudogo
WJWolfgang Jacquet
Wolfgang Jacquet
Background: Real-time outbreak analyses are often requested before surveillance systems have stabilised or epidemics have generated enough information for the desired inference. Existing approaches address surveillance quality, forecasting, estimands and identifiability separately, but do not provide a common rule for deciding which analytical product is supportable at a particular data vintage. We developed an estimand-first framework for analytical readiness. Methods: The framework distinguishes surveillance maturity (S), epidemic-process informativeness (E) and estimand-specific analytical readiness (Rθ,v), defined as whether the available data vintage, observation process, method and decision-matched validation support a specified inference for a specified decision. We stress-tested four implications using longitudinal data from the 2018–2020 Ebola response in eastern Democratic Republic of the Congo (DRC), archived geographic forecasts, independent forecasting data from Western Area, Sierra Leone, and a targeted mortality-identifiability experiment. Results: During a documented DRC surveillance disruption and recovery, seven-day persistence forecasts had all-health-zone absolute errors of 1, 3, 18 and 4 cases across pre-shock, acute-shock, early-recovery and recovery origins; the largest error occurred during early recovery. Four-week reported-case trend multipliers changed from 0.67 and 0.73 to 1.12 and 1.29, while the final fit was strongly overdispersed (Pearson dispersion 7.65), demonstrating asynchronous readiness across estimands. Archived geographic forecasts improved a Top-3 allocation decision over cumulative burden at only one origin despite consistently lower Brier scores for one specification. In Western Area, persistence forecast mean absolute error increased from 39.1 cases at one week to 82.3 at four weeks, and a history-to-horizon ratio did not define a universal threshold. An observed reported case-fatality ratio of 0.40 was compatible with constructed latent fatality values from 0.10 to 0.80; increasing the reported denominator narrowed sampling uncertainty without reducing structural uncertainty. Conclusions: Analytical readiness is task- and vintage-specific rather than a property of a dataset. More data, model convergence or narrow intervals cannot substitute for estimand definition, observation-process awareness, decision-matched validation and explicit identification analysis.

22 September 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
JVJohan G.L. GL Verheyden
Johan G.L. GL Verheyden
CMCelestin Nzanzu Mudogo
Celestin Nzanzu Mudogo
During an outbreak, analysts routinely make two judgment calls that are rarely made explicit: whether the surveillance system generating case data can be trusted, and whether the data are yet sufficient for formal calculation and prediction. Existing frameworks address adjacent problems but none formalizes the within-outbreak staging these calls represent. We propose surveillance maturity and epidemiological maturity as distinct, nested constructs. The former is a property of surveillance capacity and process, evidenced by, but not defined by, stabilization of the case-data stream. The latter is conditional on surveillance maturity plus an independent requirement that the epidemic process itself has generated sufficient signal for model parameters to be identifiable. We ground the framework in structure-process-outcome and logic-model traditions and illustrate it using two historical Ebola outbreaks and the ongoing 2026 Bundibugyo virus disease outbreak, which also demonstrates the framework's limits under sparse field evidence.